What this shows
Every day from 1980 to 2024, we ask which people were living inside a large, long-lasting drought, and add it up. The result is mapped here at ~55 km resolution, per country, and over any time span you choose.
It is an exposure account — people located inside a drought — measured before any question of who is harmed or how badly.
The key measure
Person-days. One person in drought for one day = one person-day.
1 million people in drought for 100 days = 100 million person-days. It captures how long as well as how many — which matters, because droughts last for months.
We also report the simpler people affected (how many people, in a given year).
What counts as a drought here
Only major, persistent events — at least ~125,000 km² (a third of Germany) and lasting at least 30 days. Small or brief droughts are deliberately left out, so every number is a lower bound.
Resolution: 0.5° (~55 km) grid cells. This is the finest the drought data allow — the map does not imply sharper detail. The last chart below measures what that limit costs.
Dry soil. The drought that parches fields, pastures and forests — what most people picture as a drought. This is our primary variable.
Low river flow and water supply. The drought you feel in rivers, reservoirs and taps, rather than in the soil. It can behave quite differently from soil drought.
Both at once, on the same day: crops failing and rivers low, so there is no dry-season irrigation to fall back on. The hardest-hit population.
A statistical split (LMDI) of the increase in person-days into a “more/longer drought” part and a “more people” part, for three ways of choosing the start and end years. Soil-moisture drought is overwhelmingly hazard-driven; runoff drought is not.
Black: what happened. Red dashed: what exposure would have been if population had stayed at 1980 (hazard alone). Blue dotted: if the drought pattern had stayed at 1980 (population alone). The red line tracks the black one — the hazard is doing the work.
Share of exposure vs share of population, ordering the world from least to most deprived. A curve on the diagonal means exposure is spread in proportion to population. Both curves sit almost on it — at this scale, drought is not concentrated on the most deprived.
Exposure counted on the 0.5° grid, compared with a finer 0.1° reference. Treating each coarse cell as simply “in or out” of drought (red) understates exposure — up to 17% in fragmented years like 1986. Keeping each cell's drought fraction (blue) almost removes the error.
Each point is one year. X-axis: the share of drought-affected population sitting in cells just under the 50% majority threshold — where the binary rule discards the whole cell. Y-axis: how badly that year's binary count undershot the truth. The two move together — fragmented drought footprints, not a fixed discount, drive the bias.